Stacked on #4243 (the `perf:client` harness) — first step of fixing the
every-100ms main-thread stutter: make the per-tick burst small before
spreading what remains across frames.
## Problem
The harness showed the main-thread burst was dominated by
`structuredClone` of the `updates` object, and the clone was dominated
by two kinds of per-tick churn that re-sent object payloads every tick:
- `gold` / `troops` / `tilesOwned` change for nearly every alive player
every tick → ~278 partial `PlayerUpdate` objects per tick (world/400
bots), ~508 on giantworldmap.
- Attack troop counts tick down every tick → whole
`outgoingAttacks`/`incomingAttacks` arrays re-cloned for every fighting
player every tick.
- `playerNameViewData` (an all-players record) was cloned every tick but
only recomputed every 30 ticks.
## Change
Three additions to the worker → main protocol (all transferable,
zero-clone):
1. **`packedPlayerUpdates`** — `[smallID, tilesOwned, gold, troops]`
float64 quads for players whose stats changed. These fields no longer
appear in `PlayerUpdate` diffs (first emissions still carry the full
snapshot). Gold is exact in a float64 (game values ≪ 2^53).
2. **`packedAttackUpdates`** — `[ownerSmallID, direction, index,
troops]` quads. Attack arrays are only resent when
membership/order/retreating changes — which is exactly the condition
that keeps the patch indexes valid (a tick either resends an array or
patches it, never both).
3. **`playerNameViewData` is now optional** — attached only on
placement-rebuild ticks (spawn ticks, first ticks, every 30th, spawn
end). The client keeps the last applied values; dead players' name
placements freeze at death (matching the previous effective behavior).
On the client, `GameView.populateFrame` now also rebuilds `names` /
`relationMatrix` / `allianceClusters` only when their inputs changed
that tick — field presence on a partial `PlayerUpdate` marks them dirty.
(`playerStatus`, nuke telegraphs, and attack rings still recompute every
tick; they're tick- or unit-dependent.)
## Results (perf:client, this machine; low-end devices ~5–20× slower)
Default run (world, 400 bots, 1800 ticks):
| stage | before | after |
|---|---|---|
| clone (serialize+deserialize) | 1.02ms | **0.09ms** |
| GameView.update | 0.62ms | **0.29ms** |
| WebGLFrameBuilder.update | 0.04ms | 0.04ms |
| **TOTAL burst mean** | **1.67ms** | **0.42ms** |
| TOTAL p99 / max | 3.47 / 10.3ms | **1.21 / 3.92ms** |
giantworldmap/600t: 2.54 → 0.68ms mean. Player update objects: 278 → 6.5
per tick (world), 508 → 12 (giant). The remaining burst is mostly tile
apply + per-tick derivations — the part that frame-spreading (next step)
addresses.
## Verification
- **Sim final hash unchanged** on all three reference configs
(`5607618202213430`, `29309648281599524`, `39945089450032050`) — no
simulation behavior change.
- **View hash unchanged** on all three configs (`942106e9`, `a3aae227`,
`cbaaf265`) — the rendered view state is provably identical
tick-for-tick, including the name-freeze semantics.
- New tests: `tests/PackedPlayerUpdates.test.ts` (drain + GameRunner
cadence), packed-channel and freeze-at-death cases in
`tests/client/view/GameView.test.ts`, `packAttackTroopDeltas` unit tests
and updated diff contract in `tests/GameUpdateUtils.test.ts` /
`tests/PlayerUpdateDiff.test.ts`.
- `npm test` (1490 tests), `eslint`, `prettier`, `tsc --noEmit` all
pass.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
## Problem
In team games, alternate view (space-hold) colored teammates' units red
(enemy color) instead of yellow (ally color). Teammates' territory
borders had the same problem.
## Root cause
`buildRelationMatrix()` in
`src/client/render/frame/derive/RelationMatrix.ts` already supports an
optional `teams` map that marks same-team pairs as `RELATION_FRIENDLY`,
but the call site in `GameView.populateFrame()` never passed it (the
companion `buildTeamMap` helper was dead code). Only explicit alliances
were marked friendly, so a teammate without a formal alliance read as
neutral — and the alt-view unit palette maps neutral to the enemy color.
## Fix
- `GameView` now tracks a `smallID → team` map, populated when each
`PlayerView` is first created (team is a static field, so once per
player is enough).
- The map is passed through to `buildRelationMatrix()`, which feeds both
the `AffiliationPalette` (unit colors) and `BorderComputePass` (border
colors).
## Testing
- New regression test in `tests/client/view/GameView.test.ts`: same-team
players are `RELATION_FRIENDLY` in `frame.relationMatrix`, cross-team
players stay neutral.
- All 36 GameView tests pass; typecheck clean.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
## Summary
The WebGL renderer was adapted from an external extension and carried a
lot of machinery this integration never uses (replay playback, its own
input/event system, a GL radial menu). This PR is two mechanical cleanup
passes with **no behavior change**: delete the dead code, then untangle
the `GameView` naming collision.
**78 files, +142 / −2,197.**
### Pass 1 — remove dead extension baggage
- **Replay/copy mode**: `FrameData.tileMode` was hard-coded `"live"`;
the copy branches in `frame/Upload.ts`, `UploadOptions` (never passed),
`applyFullFrame`/`applyFullTiles`/`applyDelta` on the facade and
`GPURenderer`, `HeatManager.resetForSeek`, and the seek-upload methods
on `TerritoryPass`/`TrailPass` were all unreachable. Also deletes
`types/Replay.ts`, `types/FrameSource.ts`, `types/GameUpdates.ts`,
`types/Game.ts` (imported only by the types barrel).
- **FrameEvents**: trimmed from 14 fields to the 3 actually populated
and read (`deadUnits`, `conquestEvents`, `bonusEvents`). The other 11
fed the extension's stats system and were never written or read here.
- **GL radial menu**: `RadialMenuPass`, its 4 shaders, and ~10 API
methods on facade + renderer had zero callers — the game uses the DOM/d3
radial menu in `hud/layers/RadialMenu.ts`. The pass was constructed and
drawn every frame for nothing.
- **Facade event system**: `GameViewEventMap` defined 10 event types
(`click`, `hover`, `scroll`, …) but only `contextrestored` was ever
emitted — input actually flows through `InputHandler` → EventBus →
controllers. Replaced the listener map with a single `onContextRestored`
callback and deleted `Events.ts`. Also fixed the stale header comment
claiming the facade handles user interaction.
- **Unused API surface**: removed ~20 facade/renderer methods with zero
callers (camera passthroughs like
`panTo`/`zoomTo`/`fitMap`/`screenToWorld`, hit-testing queries, SAM
replay setters, `setSelectedUnit`, `clearFx`/`setFxTimeFn`,
`onFrame`/`afterRender`/fps tracking).
Deliberately left alone: `Camera`'s pan/zoom primitives (building blocks
for a possible future camera unification) and the `timeFn` plumbing
inside the FX passes (deeply embedded as defaults; only the dead
renderer-level wrappers were removed).
### Pass 2 — untangle the three GameViews
- `render/gl/GameView.ts` → **`MapRenderer.ts`** (class `MapRenderer`).
Every importer was already aliasing it as `WebGLGameView` to dodge the
collision with the simulation-mirror `GameView` in `client/view/`, so
this removes aliasing rather than adding churn. `render/CLAUDE.md`
updated.
- Deleted the `src/core/game/GameView.ts` back-compat shim (its own TODO
asked for this). All 51 importers now import from `src/client/view/`
directly via a new 3-line barrel `view/index.ts`.
## Test plan
- `tsc --noEmit` clean, `eslint` clean
- Full test suite passes (1,385 + 65 server tests)
- Manual verification via headless Chromium: started a singleplayer game
and confirmed the renderer works end-to-end — terrain draws, spawn-phase
overlay shows, territories fill with borders after spawning, player
names/flags render, no renderer console errors
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
PlayerState.embargoes was string[] of stringified smallIDs — the
renderer parsed each entry with parseInt() to use as an array index.
Flagged in the integration handoff as something that should be number[].
Switch to number[] end-to-end: renderer type, relation-matrix derive
(no parseInt), PlayerView.setEmbargoSmallIDs / hasEmbargoAgainst
(numeric Array.includes, no String() temporaries), and GameView's
embargo translation pass. Also updates the PlayerView test that pinned
the old format.
PlayerView/UnitView now wrap renderer-shaped state objects (PlayerState,
PlayerStatic, UnitState) directly instead of holding engine wire types.
GameView owns a long-lived FrameData object kept in sync each tick:
players/units/tiles/trail/railroad are mutated in place; derived buffers
(playerStatus, relationMatrix, allianceClusters, nukeTelegraphs,
attackRings) and events are recomputed in a final populateFrame() pass.
The renderer reads gameView.frameData() and the same byte-identical
state objects PlayerView/UnitView wrap. WebGLFrameBuilder shrinks from
~270 to ~70 LOC: palette management + a single uploadFrameData() call,
no per-frame UnitState allocation on the hot path.
Wiring: maxPlayers=1024 on RendererConfig (pre-sizes NamePass/palette/
relation matrix textures); NamePass disabled so HTML NameLayer remains
the only on-screen player names.
Also: 39 new tests covering PlayerView/GameView/FrameData behavior;
replace .data field access in three layer call sites with accessor
methods (betrayals(), type(), getTraitorRemainingTicks()).
Evan